Forest Ecology and Management 256 (2008) 209–220

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

Impact of bark beetle (Ips typographus L.) disturbance on timber production and carbon sequestration in different management strategies under climate change Rupert Seidl *, Werner Rammer, Dietmar Ja¨ger, Manfred J. Lexer Institute of Silviculture, Department of Forest and Soil Sciences, University of Natural Resources and Applied Life Sciences (BOKU) Vienna, Peter Jordan Straße 82, 1190 Vienna, Austria

A R T I C L E I N F O

A B S T R A C T

Keywords: Forest management Carbon storage Natural disturbance Bark beetle Ips typographus Climatic change Secondary coniferous forests Picea abies PICUS Simulation

The likely environmental changes throughout the next century have the potential to strongly alter forest disturbance regimes which may heavily affect forest functions as well as forest management. Forest stands already poorly adapted to current environmental conditions, such as secondary Norway spruce (Picea abies (L.) Karst.) forests outside their natural range, are expected to be particularly prone to such risks. By means of a simulation study, a secondary Norway spruce forest management unit in Austria was studied under conditions of climatic change with regard to effects of bark beetle disturbance on timber production and carbon sequestration over a time period of 100 years. The modified patch model PICUS v1.41, including a submodule of bark beetle-induced tree mortality, was employed to assess four alternative management strategies: (a) Norway spruce age-class forestry, (b) Norway spruce continuous cover forestry, (c) conversion to mixed species stands, and (d) no management. Two sets of simulations were investigated, one without the consideration of biotic disturbances, the other including possible bark beetle damages. Simulations were conducted for a de-trended baseline climate (1961–1990) as well as for two transient climate change scenarios featuring a distinct increase in temperature. The main objectives were to: (i) estimate the effects of bark beetle damage on timber production and carbon (C) sequestration under climate change; (ii) assess the effects of disregarding bark beetle disturbance in the analysis. Results indicated a strong increase in bark beetle damage under climate change scenarios (up to +219% in terms of timber volume losses) compared to the baseline climate scenario. Furthermore, distinct differences were revealed between the studied management strategies, pointing at considerably lower amounts of salvage in the conversion strategy. In terms of C storage, increased biotic disturbances under climate change reduced C storage in the actively managed strategies (up to 41.0 tC ha1) over the 100year simulation period, whereas in the unmanaged control variant some scenarios even resulted in increased C sequestration due to a stand density effect. Comparing the simulation series with and without bark beetle disturbances the main findings were: (i) forest C storage was higher in all actively managed strategies under climate change, when biotic disturbances were disregarded (up to +31.6 tC ha1 over 100 years); and (ii) in the undisturbed, unmanaged variant C sequestration was lower compared to the simulations with bark beetle disturbance (up to 69.9 tC ha1 over 100 years). The study highlights the importance of including the full range of ecosystem-specific disturbances by isolating the effect of one important agent on timber production and C sequestration. ß 2008 Elsevier B.V. All rights reserved.

1. Introduction Natural disturbances are inherent key processes of forest ecosystems and a major driver of forest development in various forest biomes (e.g., Peltzer et al., 2000; Gromtsev, 2002; Payette and Delwaide, 2003; Lorimer and White, 2003; Harcombe et al.,

* Corresponding author. Tel.: +43 1 47654 4069. E-mail address: [email protected] (R. Seidl). 0378-1127/$ – see front matter ß 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.foreco.2008.04.002

2004; Splechtna and Gratzer, 2005). Disturbances may affect forests over various spatial and temporal scales, from the level of plant functional elements to landscape scale (e.g., Ulanova, 2000; Lundquist and Beatty, 2002; Splechtna et al., 2005). Important processes in natural forest development, including regeneration dynamics or inter- and intra-species competition, are strongly affected by the prevailing disturbance regime. Traditionally, forest management tries to minimize natural disturbances and related forest dynamics in favor of a deterministic sequence of regeneration, thinning and harvesting activities.

210

R. Seidl et al. / Forest Ecology and Management 256 (2008) 209–220

Schelhaas et al. (2003) in their review of historic disturbance events in European forestry clearly demonstrated the inability of forest management to completely control disturbances in forest ecosystems. According to their findings, on average over several decades, approximately 10% of the annual harvests in Europe have been salvage operations following natural disturbances. Recently, there has been a shift in management paradigms from sustainable timber production to sustainable forest management (SFM) in a much broader sense (e.g., Ministerial Conference on the Protection of Forests in Europe (MCPFE) 1993, 1998). This, together with an increased understanding of ecosystem processes, has promoted an alternative view on natural disturbances, and mimicking natural disturbance regimes is proposed as key component of an ecosystem oriented management approach (e.g., Niemela¨, 1999; Franklin et al., 2002; Palik et al., 2002). Both management approaches rely on a proper understanding of disturbance processes and the way these may interact with management. Thus, the development and application of new tools for risk assessment and forest management decision support have attracted substantial attention recently. While numerous risk assessment models are available which are designed to assess a particular state of the forest for its vulnerability to a disturbance agent in a static approach (e.g., Peltola et al., 1999; Gardiner and Quine, 2000; Gardiner et al., 2000; Gan, 2004; Achim et al., 2005; Netherer and Nopp-Mayr, 2005; Olofsson and Blennow, 2005), fewer examples exist (e.g., Lexer and Ho¨nninger, 1998; Kurz et al., 2000; Mailly et al., 2000; Keane et al., 2004; Crookston and Dixon, 2005) where disturbance agents were explicitly included in dynamic forest ecosystem models. Discussions on a likely climate change increase the need to address climate dependencies of disturbance agents explicitly in decision support tools and models (Ayres and Lombardero, 2000; Peterson, 2000; Volney and Fleming, 2000; Harrington et al., 2001; Bale et al., 2002). Dale et al. (2000) conclude that a dynamic integrative approach explicitly addressing the multiple interactions between environmental changes, forest management and disturbance agents is urgently needed to support forest resource managers. Some of the few examples attempting to address these needs were reported by Keane et al. (1996), Chen et al. (2000) and Thornton et al. (2002). Recently, increased interest has been paid to these questions, since countries can elect to account for forest management based carbon (C) sinks under the Kyoto Protocol (UNFCCC, 1997). Disturbance regimes, as influenced by a changing climate, may impair sink enhancement strategies in forestry (Breshears and Allen, 2002). So far several studies have addressed this issue, mainly focusing on the boreal zone: for fire disturbance, e.g., Seely et al. (2002), Thornley and Cannell (2004) and Ito (2005); for fire and insect disturbance, e.g., Kurz and Apps (1994), Kurz et al. (1998) and Li et al. (2003). Recently, Thu¨rig et al. (2005) presented an analysis of windthrow effects on the C balance of Swiss forests. Most of these analyses assumed stand-replacing disturbances, which directly affect the age-class structure of the simulated forests. However, analyses including explicitly the inter-related effects of climate change and disturbances operating at low to intermediate intensity levels on timber production and the C budget of forest ecosystems have rarely been presented. In Central Europe, Norway spruce (Picea abies (L.) Karst.) has been heavily promoted due to its superior productivity and ease of management outside its natural range, on sites naturally supporting broadleaved species compositions. These secondary coniferous forests are particularly prone to an array of insect and disease organisms (e.g., Klimo et al., 2000; Spiecker et al., 2004). Among these agents the spruce bark beetle Ips typographus (Scol. Col. L.) is regarded as the most important biotic risk agent for Norway spruce (Christiansen and Bakke, 1988; Schelhaas et al.,

2003). Under a warmer and possibly drier climate the vulnerability of these forests to bark beetle infestations is expected to increase drastically. Possible forest management strategies to mitigate the risk of management are discussed intensively in forest practice as well as in the scientific literature (e.g., Klimo et al., 2000; von Teuffel et al., 2005) and include transformation of current age class Norway spruce forests to continuous cover forestry (e.g., Reininger, 2000; Pommerening and Murphy, 2004; Loewenstein, 2005) as well as conversion to mixed species stands which are better adapted to the prevailing site conditions (e.g., Spiecker et al., 2004). Currently, at the operational level of a forest management unit, there is no detailed analysis of how such proposed management strategies may affect timber production and C sequestration under conditions of climatic change that explicitly takes into account natural disturbances by bark beetles. The objectives of this study were to: (i) investigate, by means of a simulation study (time horizon 100 years), the interaction of bark beetle (I. typographus) induced disturbances and alternative forest management strategies on C sequestration and timber production under conditions of climatic change; (ii) estimate the potential error of ignoring the biotic disturbance agent in the assessment, and thus contribute to the discussion on the influence of disturbances on forest functions under climate change. To this end, four alternative forest management strategies were investigated at the forest management unit level under a baseline climate scenario (detrended climate 1961–1990) and two transient climate change scenarios with the model PICUS v1.41 (e.g., Seidl et al., 2005). 2. Materials and methods 2.1. Study material 2.1.1. The study site The study was conducted for a 248.7 ha forest management unit (FMU) in the province of Carinthia in southern Austria (Lat. E14.37, Long. N46.78). The FMU is situated in the submontane vegetation belt at about 550 m above sea level. The climate regime is subcontinental. Soil conditions are characterized by mainly crystalline bedrock consisting of glacial residues with occasional calcite cliffs. The region has been managed intensively for centuries. Current forests are dominated by Norway spruce. The potential natural vegetation is dominated by deciduous species (mainly beech, Fagus sylvatica L. and oak, Quercus robur L.) with admixed Scots pine (Pinus sylvestris L.), and is mainly differentiated by soil water regime (see Mayer, 1985; Kilian et al., 1994). In general geomorphological conditions are smooth and allow for a fully mechanized harvesting technology. 2.1.2. Soil data Soil conditions of the FMU are mainly characterized by fertile Cambisols with medium to good water holding properties. This soil type dominates 98.8% of the forest area (Steiner, 1998). For the simulations, two site types were distinguished within that soil type, differing with regard to water holding capacity (ST1 and ST2). The third site type is characterized by shallow rendzic Leptosols (ST3), which prevail in a limited area at the southern border of the FMU. Data for initialization of soil C and N pools were collected from soil samples in all three site types. Mineral soil properties (see Table 1) were kept constant within a site type, whereas stand-level forest floor C and N pools were derived from statistical relationships with stand characteristics (see Seidl et al., 2007a for details). 2.1.3. Stand data Stand data were available for 103 compartments from a full inventory of the FMU (Unegg, 1999). Norway spruce is the

R. Seidl et al. / Forest Ecology and Management 256 (2008) 209–220 Table 1 Properties of the mineral soil for the site types (ST1-3) as derived from the laboratory analysis of 20 soil samples

pH (H2O) WHC (mm) C (t ha1) N (t ha–1) Soil depth (mm)

ST1

ST2

ST3

Eutric Cambisola

Eutric Cambisola

Rendzic Leptosol a

4.3 161 82.0 5.85 800

4.1 120 95.8 5.92 800

6.2 60 66.3 3.35 400

WHC: water holding capacity. a Soil type.

dominant species (93% of basal area), followed by Scots pine (6%) and a minor share of deciduous species (Q. robur, Acer spp., Salix spp.). The age-class distribution is strongly skewed towards intermediate development stages with only a small number of stands over 70 years. For the simulation, stands were grouped by cluster analysis (method: partitioning around medoids, R Development Core Team, 2006) into 25 stand types according to stand age, inventory estimates of mean annual increment over 100 years (MAI100), stocking density relative to yield table basal area (Marschall, 1975), share of Norway spruce, and basal area weighted mean diameter. A representative stand per cluster was chosen for simulation—i.e. 25 representative stand types were simulated for the FMU. Since deadwood stocks had not been recorded in the inventory they were initialized as zero. Main forest characteristics of the FMU are shown in Fig. 1 (see Seidl et al., 2007a for details). 2.2. Climate scenarios and management strategies 2.2.1. Climate scenarios Three climate scenarios were employed. Climate scenario C1 is a synthetic climate baseline and features a de-trended 100-year

211

climate series based on observed climate data of the period 1961– 1990 (mean annual temperature 7.6 8C, mean annual precipitation 1013 mm) which had been interpolated from nearby weather stations of the Austrian weather service. This generic approach was chosen in order to obtain a baseline for the assessment of changes in bark beetle infestations not superimposed by trends in climate input. The other two climate scenarios implied a transient climate change and were based on global circulation model (GCM) simulations. Both scenarios refer to the IS92a scenario (‘‘business as usual’’) of the IPCC (1995), assuming a doubling of the atmospheric CO2-concentration over the 21st century. Climate scenario C2 is based on the GCM ECHAM4-OPYC3 (European Center Hamburg, Germany) (cooling effect of sulphur particles not included), and climate scenario C3 on simulations of the GCM HadCM2 (Hadley Center, UK). Spatial interpolation of the GCM climate parameter anomalies to the study site was done using Delaunay triangulation. The climate data were provided by the Potsdam Institute for Climate Impact Research (see Kelloma¨ki et al., 2005). The climate change scenarios (C2 and C3) show steady warming until the end of the 21st century (+3.7 8C and +3.1 8C for C2 and C3, respectively for 2090–2100, relative to the first decade of the century, see Fig. 2). Precipitation patterns in the climate change scenarios show no clear trend: the scenarios predict an increase in precipitation for the first decades of the century with a decreasing trend towards the end of the 100-year period (Fig. 2). The last 20 years are particularly dry under scenario C3. 2.2.2. Silvicultural management strategies All management interventions within the strategies were specified as percent volume removals in five relative diameter classes and were individually tailored to each simulated stand type (e.g., adopted to initial conditions of stand type). Merchantable dead trees (dbh >10 cm) were removed in the year of death, inter alia mimicking current forest protection routines. Pro-active forest

Fig. 1. Summary statistics of the 103 compartments of the FMU inventory. MAI refers to mean annual increment at age 100 (gross volume over bark, m3 ha1 a1) and stocking density is relative to the yield tables of Marschall (1975). BA = stand basal area.

212

R. Seidl et al. / Forest Ecology and Management 256 (2008) 209–220

Fig. 2. Mean annual temperature (a) and precipitation (b) in the climate scenarios. C1: de-trended baseline climate based on the period 1961–1990. C2: climate change scenario for the IS92a storyline based on scenario output of ECHAM4-OPYC3. C3: climate change scenario for the IS92a storyline based on scenario output of HadCM2. See text for details.

protection measures were not simulated, but mortality-induced changes in stand characteristics had an impact on subsequent regular harvest removals, since harvest intensity was specified in percent of standing volume stock. Thus, a partial trade-off between regular harvests and salvage was considered with this study design. Four silvicultural management options were investigated (see Table 2). Strategy MS1 is an age-class system and reflects business as usual management with a rotation period of 90 years. Norway spruce is regenerated naturally by a shelterwood approach (regeneration period approximately 10 years), followed by a pre-commercial thinning to control for stand density. The subsequent thinning regime is characterized by several selective thinnings (Scha¨delin, 1942; Johann, 1987). Alternative management strategies were based on recently discussed silvicultural alternatives for secondary Norway spruce forests in Central Europe (see Spiecker et al., 2004). Strategy MS2 is motivated by the discussion of potential benefits of improved vertical forest structure and represents a continuous cover management regime with natural regeneration of Norway spruce (see Reininger, 2000; Pommerening and Murphy, 2004). This strategy implies several thinnings from above (‘‘structural thinnings’’) in early development phases to promote differentiation of stand structure, and a subsequent transition to a target diameter harvesting regime. Another recently discussed silvicultural option especially relevant for secondary coniferous forests is the conversion to mixed stands with broadleaved species better adapted to prevailing site conditions (MS3). In MS3 the rotation period for current Norway

spruce stands is reduced to 80 years, and beech and oak are introduced. Within MS3 several conversion variants are distinguished according to initial stand and site conditions, ranging from conversion to pure oak (21.1% of FMU area) and beech (4.5% of FMU area) after clear-cutting of Norway spruce to mixed stands of

Table 2 Characterization of the applied management strategies (MS1–4) Strategy

Management intervention

Scheduled removal (% of volume)

MS1

Pre-commercial thinning (20) Thinning from above (40, 50, 60) Thinning from below (70) Shelterwood cut (80) Clearcut (90)

40–40–40–30–30 00–10–10–25–15 00–10–10–10–05 00–25–30–35–10 100–100–100–100–100

MS2

Structural thinning (40, 50, 60) Target diameter harvest (15-year intervals starting at age 75)

00–10–15–30–20 00–00–00–10–75

MS3

Pre-commercial thinning (20) Thinning from above (40, 50, 60) Thinning from below (70) Clearcut (80)

40–40–40–30–30 00–10–10–25–15 00–10–10–10–05 100–100–100–100–100

MS4

None

Numbers in parenthesis give the approximate stand age of the respective management intervention, removals are specified as percent volume removals in five relative diameter classes (lowest to highest diameter class from left to right). Values for MS3 relate to the current coniferous forests and are adapted accordingly to the successively changed species composition. See text for details.

R. Seidl et al. / Forest Ecology and Management 256 (2008) 209–220

213

Table 3 Stand conversion variants within management strategy MS3 Conversion variant

Target share of deciduous species (%)

Percent of forest area (%)

Silvicultural characteristics

Oak

100

21.1

Planting after clearcut

Beech

100

4.5

Planting after clearcut

Spruce-beech (1)

30

68.6

Spruce-beech (2)

20

5.8

Underplanting of beech in groups/gaps under spruce canopy, natural regeneration of spruce Introduction of beech in groups in already existing natural regeneration of spruce

Spruce-beech variant (2) was applied in stands with existing natural regeneration of Norway spruce.

spruce and beech established by underplanting of beech under Norway spruce shelter and subsequent natural regeneration of spruce (74.4% of FMU area). A detailed description of the conversion variants can be found in Table 3. A ‘‘do nothing’’ (i.e., no active management) strategy (MS4) was simulated as a reference scenario with natural Norway spruce regeneration establishing spontaneously, and dead trees remaining in the forest. 2.3. Model approach and simulation design 2.3.1. PICUS v1.41 The model PICUS v1.41 used for the assessment builds on the hybrid forest patch model PICUS v1.31 (Seidl et al., 2005). The hybrid model approach aims at combining the strengths of both, patch models and process based production models, while circumventing the limitations of the individual approaches (see Ma¨kela¨ et al., 2000). Spatial basis of the simulation approach is a 10 m  10 m patch array extended into the third dimension by 5 m crown cells. In contrast to classical patch models (compare Shugart, 1984; Botkin, 1993) spatial interactions between the patches are taken into account by simulating a detailed three-dimensional light regime and spatially explicit seed dispersal. Inter- and intra-species competition is modeled based on the patch model approach presented by Lexer and Ho¨nninger (2001) whereas stand level net primary production is derived according to the simplified physiological principles of radiation use efficiency of the 3-PG model (Landsberg and Waring, 1997). The coupling of the two modeling approaches is accomplished inter alia via the stand level leaf area and is described in detail in Seidl et al. (2005). The model requires monthly input of temperature, precipitation, radiation and vapor pressure deficit. In extension to model variant 1.31 two additional modules have been integrated. First, a process-based soil model of dynamic C and N cycling (Currie et al., 1999; Currie and Nadelhoffer, 1999) has been added (Seidl et al., 2007a). Interaction of aboveground production processes and belowground C and N dynamics are simulated on a monthly time-step. Soil organic C and N microbialdetrital pools, implicitly containing microbial biomass, are simulated for the forest floor and mineral soil. Fine litter enters detrital pools according to different C-classes and processes of nitrogen immobilization and mineralization are simulated. Furthermore, mass and N dynamics of fine and coarse woody debris are considered, including humification. The model explicitly simulates NH4+ and NO3 pools and fluxes and tracks leaching fluxes between the soil horizons. The general concept has been thoroughly tested in several studies (Currie et al., 1999; Currie and Nadelhoffer, 1999; Moorhead et al., 1999; Currie et al., 2004). Motivated by the high relevance of biotic disturbances for forest management in Central Europe, an earlier approach to model bark beetle disturbances by Lexer and Ho¨nninger (1998) was adopted and improved by new findings on the physiology and development of I. typographus (e.g., Netherer and Pennerstorfer, 2001; Netherer

and Nopp-Mayr, 2005; Baier et al., 2007). The bark beetle disturbance module includes (i) the stochastic computation of the infestation risk for simulated forest stands, (ii) the estimation of damage intensity if an infestation occurs, and (iii) the spatial distribution of tree mortality in the stand. Damage risk is derived from the number of potential beetle generations per year estimated using a thermo-energetic model approach (Coeln et al., 1996; Wermelinger and Seifert, 1999; Netherer and Pennerstorfer, 2001; Baier et al., 2007): annual potential beetle generations are calculated according to thermal requirements represented by a sum of degree-days (557) above a threshold temperature for beetle development (8.3 8C). Bark temperatures above the development optimum (30.4 8C) lead to a slowing and finally complete halt (>38.9 8C) of beetle development. Swarming requirements in terms of a combined day length and air temperature threshold are introduced for the start of a new generation. Currently, the simulation of potential generations does not account for a perennial bark beetle gradation – i.e. beetle development starts from zero every year. Annual potential generations are taken as a proxy of thermal environmental conditions for beetle development and are transformed into a stand level hazard rating introduced by (cf. Lexer (1995), Netherer and Nopp-Mayr (2005)). In addition, a model of stand predisposition to I. typographus infestations is adopted from Lexer (1995) and Netherer and NoppMayr (2005), including four stand level predisposition indicators (share of host trees in a stand, stand density (basal area), stand age, and Norway spruce drought index). Stand predisposition and thermal predisposition (potential generation score) are used to fit an empirical function to data on bark beetle infestations from Lexer (1995) to derive an annual probability for bark beetle damage (Eq. (1)). This simulated damage probability is assessed against a random number to derive whether damage from bark beetle occurs in a given year of the simulation. pBByr ¼ 1  eðx1PI

x2 GEN

Þ

(1)

pBByr = annual probability for bark beetle damage; PI = stand predisposition index to bark beetle damage [0, . . ., 1]; GEN = thermal predisposition scoring for potential generation number [0, . . ., 1]; x1 = empirical coefficient (1.51); x2 = empirical coefficient (1.65). Damage intensity is calculated according to the empirical findings of Lexer (1995) using a stand hazard index combining a rating of south- and east-exposed stand edges, drought stress, and proportion of Norway spruce host trees in the stand (see Lexer, 1995) (Eq. (2)). Drel ¼

1 1 þ ex3x4SHIyr

(2)

214

R. Seidl et al. / Forest Ecology and Management 256 (2008) 209–220

Drel = annually damaged stems in a stand [share of stem number]; SHIyr = annual stand hazard index; x3 = empirical coefficient (3.9725); x4 = empirical coefficient (2.9673). The spatial allocation of damage is derived from a predisposition ranking of the patches within the simulated stand and assumptions on the extent of an outbreak spot. A more detailed description alongside a thorough sensitivity analysis can be found in Seidl et al. (2007b): Model behavior along an elevation gradient in the Eastern Alps in Austria was found to be realistic and compared well with general observations of forest pest monitoring systems (Krehan et al., 2005). PICUS v1.41 also contains a flexible management module allowing for spatially explicit harvesting and planting operations at the individual tree level. PICUS has been successfully evaluated with regard to the simulation of natural forest succession and equilibrium species composition over broad environmental gradients in the Eastern Alps. Furthermore, in a comparison with long-term growth and yield data the model was found capable of reproducing volume production and stand structure of managed, uneven-aged multi-species stands (Seidl et al., 2005). Table 4 summarizes the main features of PICUS v1.41 following the scheme of Freeman et al. (2005), who compared the structure and function of six models applied for the simulation of forest management under climate change in Europe.

stand type (total number: 103). This design granted comparability of the two simulation series UD and BB by applying the same 25 initial stand types, and also ensured a realistic consideration of disturbance risk, proportional to the number of stands represented by a stand type. The disturbance regime was characterized in terms of annual infestation frequency and damage intensity. Infestation frequency for the whole FMU was expressed as the proportion of stands with bark beetle infestation in a respective year. Damage intensity was presented as relative number of trees damaged annually per stand. A mean intensity over all damaged stands was calculated per year. Timber production was assessed in terms of harvests, share of salvage from bark beetle infestations, and standing volume at the end of the simulation period. Timber volume is merchantable volume under bark, except where stated otherwise. C cycle effects were evaluated applying a net stock change approach, e.g. a summation of annual stock changes over the simulation period. Analyzed compartments were living tree C (including stem-, branch-, leaf- and root-carbon), C stored in deadwood (standing deadwood and downed woody debris of >10 cm in diameter), and soil C (consisting of soil organic carbon, litter, humified matter and fine woody debris